616
JOURNAL OF PHYSICAL OCEANOGRAPHY
VOLUME 31
Impact of the Midlatitude Storm Track on the Upper Pacific Ocean
WILCO HAZELEGER
Royal Netherlands Meteorological Institute, De Bilt, Netherlands
RICHARD SEAGER, MARTIN VISBECK,
AND
NAOMI NAIK
Lamont-Doherty Earth Observatory of Columbia University, Palisades, New York
KEITH RODGERS
Max-Planck-Institut für Meteorologie, Hamburg, Germany
(Manuscript received 19 November 1999, in final form 5 June 2000)
ABSTRACT
Transient eddies in the atmosphere induce a poleward transport of heat and moisture. A moist static energy
budget of the surface layer is determined from the NCEP reanalysis data to evaluate the impact of the storm
track. It is found that the transient eddies induce a cooling and drying of the surface layer with a monthly mean
maximum of 60 W m22 . The cooling in the midlatitudes extends zonally over the entire basin. The impact of
this cooling and drying on surface heat fluxes, sea surface temperature (SST), water mass transformation, and
vertical structure of the Pacific is investigated using an ocean model coupled to an atmospheric mixed layer
model. The cooling by atmospheric storms is represented by adding an eddy-induced transfer velocity to the
mean velocity in an atmospheric mixed layer model. This is based on a parameterization of tracer transport by
eddies in the ocean. When the atmospheric mixed layer model is coupled to an ocean model, realistic SSTs are
simulated. The SST is up to 3 K lower due to the cooling by storms. The additional cooling leads to enhanced
transformation rates of water masses in the midlatitudes. The enhanced shallow overturning cells affect even
tropical regions. Together with realistic SST and deep winter mixed layer depths, this leads to formation of
homogeneous water masses in the upper North Pacific, in accordance to observations.
1. Introduction
The sea surface temperature (SST) forms the lower
boundary condition for the atmosphere. Therefore, understanding of how the SST and its variability is generated is of prime importance for climate studies. The
SST is modified by surface heat fluxes. Variability in
the surface fluxes as well as changes in the ocean heat
flux divergence drives variability in SST (e.g., Cayan
1992). This makes a good understanding of heat fluxes
essential for climate studies. Also, from an oceanographic viewpoint surface heat fluxes are important.
Heat fluxes drive water from one density to another in
the mixed layer, leading to transformation of water masses (Walin 1982). Horizontal and temporal variations in
the surface fluxes lead to variations in the mixed layer
depth. This determines the amount of water that is subducted to the interior. Water mass transformation and
Corresponding author address: Dr. Wilco Hazeleger, Royal Netherlands Meteorological Institute (KNMI), P.O. Box 201, 3730 AE,
De Bilt, Netherlands.
E-mail: [email protected]
q 2001 American Meteorological Society
subduction determine the stratification of the ocean
(Woods 1985; Marshall et al. 1998). The importance of
the surface heat fluxes to the mean state and the variability of the ocean requires a good understanding of
the fluxes and a good representation of them in numerical models. In this manuscript we study one aspect
of the surface fluxes in the Pacific, that is, the effect of
storm tracks on the heat fluxes and subsequently the
effect on SST, water mass transformation, and the stratification of the ocean. The purpose of this paper is twofold. First, we diagnose the additional cooling due to
atmospheric storms and determine an eddy-induced
transfer velocity that can be used in atmospheric mixed
layer models to account for the effect of storms. Then,
the eddy-induced transfer velocity will be applied in a
numerical model to study the impact of the storm tracks
on the upper Pacific Ocean.
In this paper we will focus on the surface heat fluxes
at the midlatitudes in the Pacific. This is a key region
for climate variability because the subduction rates are
large (Qiu and Huang 1995) and a positive ocean–atmosphere feedback may occur in this region (Latif and
Barnett 1994). In this region transient atmospheric ed-
FEBRUARY 2001
617
HAZELEGER ET AL.
dies form into localized, geographically fixed areas.
These so-called storm tracks are located near the midlatitude atmospheric jet streams and, interestingly, also
follow the regions of maximum meridional SST gradient. Here, the poleward and upward sloping isentropes
provide available potential energy for the baroclinic instability process. The flattening of isentropes by baroclinic instability cause a transport of warm and moist
air northward and cool dry air southward (Kushner and
Held 1998). The northward sensible and latent heat
transport in the atmosphere reflects this eddy-driven circulation. It cools and dries the surface layer at the midlatitudes (e.g., Valdes and Hoskins 1989). The impact
of this cooling on the ocean is studied in this paper.
Earlier studies with one-dimensional models by, for example, Camp and Eslberry (1978) have shown that the
effect of individual storms on the oceanic mixed layer
is large. They also indicate that nonlocal processes may
be important. In a very idealized ocean model Polonsky
et al. (1992) show a cooling of 1–2 K due to synoptic
atmospheric variability. They emphasize the importance
of horizontal oceanic inhomogeneity. Here we will use
a much more elaborate three-dimensional ocean model
to study the impact of storms.
Synoptic variability associated with the atmospheric
transient eddies has a timescale of 2–10 days (Wallace
et al. 1988). Alexander and Scott (1997) have studied
the submonthly variability in the surface buoyancy fluxes. They showed that more than half of the total surface
flux variability in winter in the region where storms are
active occurs on timescales of less than 30 days. Variability on these short timescales is not always included
in surface climatologies or simple models used to force
ocean models (e.g., Seager et al. 1995; Kleeman and
Power 1995). This requires parameterization of the submonthly variability. Here, we present a parameterization
of the impact of storm tracks on the surface fluxes based
on the parameterization of eddies in the ocean according
to Gent and McWilliams (1990). That is, an eddy-induced advection is used to parameterize the effect of
eddies on temperature and humidity. Kushner and Held
(1998) have shown that such an approach is applicable
to the atmosphere. First, we will present a moist static
energy budget of the atmospheric surface layer to clarify
the spatial and temporal structure of the eddy fluxes.
Then the parameterization will be tested in an atmospheric mixed layer model. In this model the turbulent
fluxes are determined through a balance of horizontal
advection and diffusion, the surface flux and the flux at
the mixed layer top, and, for temperature, radiative cooling. The latent heat, sensible heat, and longwave radiation fluxes can evolve freely. Furthermore, the temperature and humidity are internally generated in the
model. Therefore, this is a suitable model to test the
effect of the cooling by storms. Instead of parameterizing the eddy-induced transfer velocity itself, we will
specify it according to the National Centers for Environmental Prediction–National Center for Atmospheric
Research (NCEP–NCAR) 40-year reanalysis data (Kalnay et al. 1996). Thus the position and strength of the
storm tracks remain constant (except for a seasonal cycle). However, the gradients upon which the additional
velocity acts are internally determined by the model.
By including the parameterization, the tendency to
moisten and warm the boundary layer through increased
ocean–atmosphere surface fluxes due to storms is also
explicitly simulated. After validation of the parameterization, it is subsequently used to study the effect of the
storm tracks on the ocean using an oceanic general circulation model.
This manuscript is organized as follows. In section 2
the moist static energy of the surface layer of the atmosphere is presented using NCEP–NCAR reanalysis
data. Then the parameterization of the impact of the
storm tracks on the surface fluxes is presented in section
3. In section 4 the models are presented, and in section
5 the impact of the storm track on the heat fluxes, SST,
water mass transformation, and the interior of the ocean.
Finally in 6 a summary is given and conclusions are
drawn.
2. Moist static energy budget of the lower
atmosphere
In order to study the effect of the atmospheric eddies
on the surface fluxes we will first present a moist static
energy budget of the lower atmosphere. We focus on
moist static energy because it is a conserved quantity
for air masses and because it combines the effects of
humidity and temperature. The budget gives insight in
the processes that determine the surface fluxes. Such a
budget for the atmospheric mixed layer has been presented by Seager et al. (2000). There, the focus was on
anomalies related to the North Atlantic Oscillation.
Here, we show the budget for the mean fields in the
Pacific.
The moist static energy is defined as
S 5 gz 1 c p u 1 L y q,
(1)
with g the gravitational constant, z the height, c p the
heat capacity of air, u the potential temperature, L y the
latent heat, and q the specific humidity. We consider a
layer between the surface (1000 hPa) and 925 hPa (the
second level of the NCEP–NCAR reanalysis data). The
layer is assumed to be in steady state. This is justified
by the fast adjustment of the near-surface air temperature to SST (Boers and Betts 1988). This is confirmed
by the high correlation between the monthly mean air
temperature and SST in the NCEP–NCAR reanalysis
data. The correlation is well above 0.6 almost everywhere in the Pacific (not shown). The vertical integral
of the moist static energy equation, in pressure coordinates, over this layer is
618
JOURNAL OF PHYSICAL OCEANOGRAPHY
1
(Pu · =S 2 v B (SB 2 S) 1 P= · (u9S9) 2 v9S9B )
g
5
v0
1
1
(S0 2 S) 1 (v 0S0) B 1 cp PR.
g
g
g
(2)
Here the subscript B stands for the top of the layer, the
subscript 0 stands for the surface; P is the pressure
thickness, u 5 (u, y ) is the wind vector, v is the vertical
pressure velocity, and R is the radiative cooling. The
primes stand for the deviation from the monthly mean.
The double primes represent the turbulent fluxes. We
analyze the terms in Eq. (2) using the NCEP–NCAR
reanalysis data from January 1958 to December 1998
(Kalnay et al. 1996). Monthly mean data has been used
with a resolution of 2.58. The last two terms in Eq. (2)
cannot be determined from the data. Their combined
effect balances the other terms and can be determined
as a residual. Note that, when the last term is diagnosed
as a residual, this term also includes errors in the calculation of the other terms. However, we expect the
errors to be small since we use the same pressure levels
as in the NCEP–NCAR reanalysis data. This excludes
one of the major sources of errors in budget studies,
that is, the vertical interpolation (see Alexander and
Schubert 1990). Also, as argued above, the tendency of
moist static energy can be omitted as it is small. Moreover, we do not focus on the residual, but rather on the
covariances in the budget, which are explicitly carried
in the NCEP–NCAR reanalysis data. The main source
of error that remains is the error in the data itself. An
estimate of that uncertainty can be obtained by making
the budget with winds and heat fluxes from the Comprehensive Ocean–Atmosphere Data Set (COADS) data
(Da Silva 1994). In this case the residual appeared to
be about 20 W m22 lower, but it has the same spatial
distribution. The differences are mainly caused by lower
surface heat fluxes in the COADS data.
Figure 1 shows the different terms of Eq. (2) in the
Northern Hemisphere winter. We only show boreal winter values here because the signals are strongest then.
Also the seasonal cycle is less strong in the Southern
Hemisphere (see section 3). When discussing the storm
track parameterization and the impact of the storms we
will also address the results from the summer season.
In Fig. 1a the horizontal advection of mean moist static
energy is shown. The advection of cool and dry air is
clear off the coast of Japan. Here northwesterly winds
advect cool and dry air from the continent. It is notable
that this advective effect is restricted to the region quite
close to the coast. To the east of this region, southwesterly winds advect warm and moist air to the north.
Equatorward of about 258N the trade winds cause a
cooling and drying of the air. The second term in Eq.
(2) represents the vertical advection of moist static energy. This term has the sign of the vertical velocity at
the top of the mixed layer (Fig. 1b). The rising and
sinking of air in the Tropics and subtropics associated
VOLUME 31
with the Hadley cell is the most prominent feature, and
elsewhere the term is small. The covariances in Eq. (2)
represent the atmospheric eddies, that is, variations on
a submonth timescale. Figures 1c and 1d show the effect
of the eddies. When the horizontal and vertical eddy
terms are added, the storm tracks become clear. The
eddies cause a cooling and drying over the midlatitudes.
The horizontal component shows that eddies transfer
moist warm air northward and cool dry air southward.
This results in a dipole pattern centered on the maximum
SST gradient (around 358N). The vertical eddy component is everywhere of the same sign and reveals a
cooling. The maximum of this cooling coincides with
the maximum SST gradient. The patterns are consistent
with results from Valdes and Hoskins (1989) and Swanson and Pierrehumbert (1997). They show a cooling by
transient eddies extending over the entire North Pacific
basin, with a maximum cooling around 1708E as well.
In Figs. 1e and 1f the right-hand side terms of Eq. (2)
are shown. Figure 1e shows the cooling by the surface
fluxes. The latent heat flux is the largest contributor to
the surface fluxes. Warming by the surface fluxes is
largest in the Kuroshio region, consistent with previous
studies (e.g., Cayan 1992). Finally, the residual term is
shown in Fig. 1f. This term must be balanced by radiative cooling and turbulent entrainment at the top of
the layer. The residual term is mainly balanced by entrainment at the top of the mixed layer, because the
radiative cooling is relatively uniform and small [O(50
W m22 )]. Essentially the same picture arises in the
Southern Hemisphere during austral winter (not shown).
In the Tropics the moist static energy budget is essentially a balance between surfaces fluxes and entrainment
at the top of the mixed layer (e.g., Seager et al. 1995).
The high outgoing surface fluxes to the east of Japan
during winter have often been attributed to advection
of cool and dry air from the continent over warm surface
waters in the western boundary currents (e.g., Cayan
1992). The budget analysis shows that horizontal advection cannot balance the warming and moistening by
surface fluxes (compare Fig. 1a with Fig. 1e) and, instead, atmospheric eddies make an equal contribution.
Furthermore, there is a large contribution from the entrainment of dry air at the top of the mixed layer that
is included in the residual term. This is most likely
indicative of active deep or shallow convection, hardly
surprising given the enormous air–sea thermodynamic
disequilibrium in this region.
The previous analysis shows that the surface fluxes
are balanced by different processes. The atmospheric
eddies are found to contribute significantly in the midlatitudes. In the next section we will present a parameterization for the impact of the eddies on the heat budget. Then the effect of the storm tracks on oceanic variables will be determined by implementing the parameterization in a simple model of the atmospheric mixed
layer.
FEBRUARY 2001
HAZELEGER ET AL.
619
FIG. 1. Spatial distribution of the contribution to moist static energy budget in the surface layer [see Eq. (2)] in W m 22 (mean over Dec,
Jan, Feb). (a) Horizontal advection, (b) subsidence, (c) horizontal eddy fluxes, (d) vertical eddy fluxes, (e) surface fluxes, (f ) residual. All
terms are plotted as if they were on the right-hand side of a moist static energy equation. The residual is the sum of the fields of (a)–(d).
3. A parameterization of the impact of storms on
the surface fluxes
In the midlatitudes, baroclinic atmospheric eddies develop as available potential energy provided by the poleward and upward sloping isentropes is converted to kinetic energy. The transient eddies accomplish a net
northward transport of heat and moisture as indicated
by a cooling of the surface layer (Figs. 1c and 1d). Here
we will use a model of the atmospheric mixed layer
coupled to an ocean general circulation model to study
the effect of the eddies on the upper ocean.
In our previous model of the atmospheric mixed layer
the effect of eddies was neglected (e.g., Seager et al.
1995). This prompted us to seek a parameterization of
the effect of the baroclinic eddies on the atmospheric
mixed layer. In a study of eddy-induced circulation in
the ocean, Gent and McWilliams (1990) showed that an
additional advection of tracers arises from the correlation of the eddy components of layer thickness and ve-
620
JOURNAL OF PHYSICAL OCEANOGRAPHY
locity. This defines an eddy-induced transport velocity.
Gent and McWilliams (1990) parameterized this eddy
transport by a downgradient Fickian diffusion of isopycnal layer thickness. This mimics the baroclinic instability process. Here we use the same ideas to parameterize the eddies in the atmosphere. Kushner and Held
(1998) have shown that this approach is applicable to
the atmospheric eddies.
The impact of eddies on the atmospheric surface layer
can be parameterized by adding an eddy-induced transfer velocity, u* to the velocity u; that is,
1
1
P(u 1 u*) · =S 2 v B (SB 2 S) 5 F.
g
g
(3)
Here F are the terms on the right-hand side of Eq. (2).
An effective eddy-induced transfer velocity can be determined from the observed eddy fluxes of moist static
energy in the lower atmosphere and the gradients of air
temperature and humidity:
u* · =S 5 = · u9S9 2 v9S9/P.
(4)
In order to diagnose u* from the eddy flux divergences using Eq. (4) a long-term synoptic dataset is
necessary. Here we use the NCEP–NCAR reanalysis
data from 1958 to 1998 to determine u* . Monthly values
of covariances in Eq. (4) and of the moist static energy
at the surface are deduced from the data to construct
monthly u* fields. The covariances of the vertical velocity and moist static energy are determined from the
925-hPa surface, which is the second level in the NCEP–
NCAR reanalysis dataset. We chose to add the vertical
eddy correlation term to the meridional eddy correlation
term. This is reasonable since the v9S9 term exchanges
warm, moist low-level air on the equatorward side of
the storm track with cool and higher dry air on the
poleward side. The y 9v9 term at 925 hPa is negative
over most of the North Pacific in the winter; that is,
anomalous southward transport is correlated with anomalous downward motion. This leads to cooling and drying of the surface air. This is analogous to a southward
horizontal transport.
In Fig. 2 we show the diagnosed u* in January and
July. As expected, the amplitude is largest in the midlatitudes, where the isentropes slope upward and poleward associated with the midlatitude jet. Here the atmospheric eddies induce a cooling and drying, which
is indicated by the equatorward directed vectors. The
u* forms into an elongated band stretching from the
northeast of Japan to Alaska and from the southeast of
Australia to Chile. It is remarkable that the spatial structure extends zonally over the entire Pacific. The observed flux divergences are higher at the western side
of the basin than at the eastern side (Figs. 1c,d). The
horizontal gradient of the mean moist static energy causes the u* to extend more zonally than the eddy flux
divergences themselves. However, as =S is smaller on
the east side, the flux divergences on the east side of
the basin are smaller than on the west side. In the North-
VOLUME 31
ern Hemisphere summer u* is greatly reduced. In the
Southern Hemisphere, the reduction of u* in the summer
is less strong. This is clearly displayed in the seasonal
variation of the zonally averaged u* displayed in Fig.
3. Note also that over the western boundary currents u*
is slightly smaller than over the open ocean. We will
address this feature later in this paper.
In the next section we will specify the diagnosed
eddy-induced transfer velocity in an atmospheric mixed
layer model in order to validate the parameterization.
In the atmospheric mixed layer model the surface heat
fluxes are internally generated. Including the u* will
enhance the ocean-to-atmosphere surface fluxes and will
tend to warm and moisten the surface boundary layer
and cool of the sea surface temperature. Note that by
using this procedure the position and strength of the
storm track is fixed (except for seasonal variations).
Low-frequency variability in the storm tracks is observed (Lau 1988). However, in this study we focus on
the mean effect of the storms and a parameterization of
that process on the atmospheric mixed layer. We do not
study the impact of the storms on mechanical stirring
and oceanic momentum fluxes. Model experiments with
daily winds instead of monthly mean winds showed no
big differences. Both model versions contained similar
SST errors and indicate that cooling and drying by the
storms need to be accounted for (see section 5). Note
that by using daily winds the horizontal eddy fluxes were
accounted for. Also the wind stresses included daily
variations. However, the vertical eddy fluxes could not
be accounted for as the vertical advection terms are
parameterized in the atmospheric mixed layer model,
which will be introduced in the next section. It appeared
that this term especially needs to be properly parameterized. The dominance of the vertical eddy fluxes of
moist static energy compared to the horizontal eddy
fluxes was already noted in section 2 (see Figs. 1c,d).
A successful representation of the cooling by storms
can be used to study the effect of the storm track on
air–sea interaction and the oceanic state. An atmospheric
mixed layer model coupled to an ocean model will be
run with and without the additional eddy-induced transfer velocity to study the impact of the storm track. The
models are described in the next section.
4. Experimental setup
a. The atmospheric mixed layer model
The atmospheric mixed layer (AML) model used in
this study simulates either a dry convective layer or a
subcloud marine boundary layer of a thickness of 60
hPa. The model is described by Seager et al. (1995) and
has been used extensively for tropical studies. In the
AML the virtual potential temperature and the specific
humidity are computed from a balance of advection,
surface fluxes, fluxes at the top of the mixed layer, and
radiative fluxes. The surface fluxes are determined by
FEBRUARY 2001
HAZELEGER ET AL.
621
FIG. 2. The eddy induced transfer velocity in (a) Jan and (b) Jul, derived from the NCEP–NCAR
reanalyses data. Transport of moist static energy by the eddy-induced velocities (in W m 22) is contoured.
familiar bulk transfer formulae. The fluxes at the top
are parameterized according to Seager et al. (1995). For
the exact formulation of the model and the parameters
therein, we refer to Seager et al. (1995) and Seager et
al. (2000). In the AML the surface latent and sensible
heat and longwave radiative fluxes can evolve freely
and air temperature and humidity variations are internally computed.
The AML is forced by monthly mean wind speed and
vector wind fields from the NCEP–NCAR reanalysis
622
JOURNAL OF PHYSICAL OCEANOGRAPHY
VOLUME 31
FIG. 3. Seasonal variation of the meridional component of the eddy-induced transfer velocity (u )
*
averaged between 1808 and 1408W (m s21 ).
data. Furthermore fractional cloudiness and the solar
radiation at the sea surface is prescribed (ISCCP, Bishop
and Rossow 1991). At the lateral boundaries the potential temperature and the specific humidity are prescribed
(NCEP–NCAR reanalysis). When the model is not coupled to the ocean, the SST is prescribed according to
Levitus and Boyer (1994).
The model domain consists of the Pacific basin reaching from 628N to 628S, 908E to 808W. Away from the
equator, the horizontal resolution is 2.58 in the meridional and zonal directions. In the Tropics the meridional
resolution is increased to 0.58. This resolution is based
on that of the ocean model, which has been chosen to
better resolve the critical scales of the equatorial waveguide.
A novelty compared to previous versions of the AML
model is the inclusion of a parameterization of the impact of the storms on the heat budget according to Eq.
4. In earlier versions of the model (e.g., Visbeck et al.
1998) the eddy-induced transfer velocity was parameterized in terms of gradients of surface air temperature.
In that model version the strength and position of the
storm track was not constrained. The parameterization
appeared not to represent the eddies very well. Also,
using daily winds did not improve the storm-track-driven heat flux. Therefore, in the present study, we chose
to specify the eddy-induced velocity derived from the
reanalysis data. We ran the AML for 10 years with pre-
scribed SST to validate the parameterization. Figures
4a,b show how the simulated eddy flux divergences of
moist static energy [left-hand side of Eq. (4)] compare
to the observed eddy flux divergences from the NCEP–
NCAR reanalysis data [right-hand side of Eq. (4)]. Note
that the moist static energy gradients are internally generated by the AML. Over most of the basin the difference between simulated and observed eddy flux divergences is less than 10 W m22 . As the amplitude of the
difference is at least five times smaller than that of the
flux divergence itself (Figs. 1c,d), the parameterization
can be considered successful. Significant differences are
only found close to the southern boundary in July. As
the parameterization represents the eddy fluxes well,
simulations with and without the eddy-induced transfer
velocity can be used to study the effect of the storm
tracks on the heat fluxes and the oceanic state (see section 5).
b. The dynamic ocean model
The numerical ocean model is a primitive equation
z-coordinate model that originates from the Gent and
Cane model (see Visbeck et al. 1998; Rodgers et al.
1999). Among the main differences compared to the
reduced-gravity Gent and Cane model is the inclusion
of a barotropic solver according to Naik et al. (1995)
together with a realistic bathymetry. Also, the model
FEBRUARY 2001
HAZELEGER ET AL.
623
FIG. 4. Difference between observed (NCEP–NCAR reanalyses) and modeled eddy induced transport of
moist static energy (u =S) in the atmospheric mixed layer (in W m22 ).
*
includes a Kraus–Turner surface mixed layer parameterization and a simple 1½-layer thermodynamic ice
model now. The Gent and McWilliams (1990) eddy parameterization is implemented with spatially varying
diffusivity according to Visbeck et al. (1997). The vertical mixing is Richardson number dependent. The mod-
el uses a Total Variation Diminishing scheme for the
horizontal advection terms. For the time differencing a
second-order Lorenz N-cycle is used. The model has 20
layers in the vertical, of which 10 are in the upper 1000
m. The horizontal resolution corresponds to that of the
AML model. At the southern edge of the domain the
624
JOURNAL OF PHYSICAL OCEANOGRAPHY
temperature and salinity are restored to observed values
from Levitus and Boyer (1994) and Levitus et al. (1994).
The barotropic transport at the Indonesian Throughflow
is set to a value of 10 Sv (Sv [ 106 m3 s21). The ocean
model is coupled to the AML model to create the socalled LOAM model.
5. The impact of midlatitude storm track on
surface heat fluxes, SST, and thermocline
a. Impact on surface heat fluxes
As shown in section 4a, the effect of storm tracks can
be successfully represented by an eddy-induced transfer
velocity derived from observed eddy flux divergences
and gradients of moist static energy. So, to study the
impact of the storms on the surface heat fluxes, the AML
model (SST prescribed) can be run with and without
the additional eddy-induced transfer velocity. Figure 5a
shows the difference between the latent and sensible
heat fluxes in both runs after 10 years. Clearly, the eddies provide an extra monthly mean cooling of up to
60 W m22 in the midlatitudes. The broad region of cooling is consistent with the pattern of the eddy transfer
velocity (Fig. 2). Note that the maximum cooling occurs
far offshore. In earlier versions of the model (e.g., Visbeck et al. 1998; Seager et al. 2000) the u* has been
related to the simulated surface air temperature. This
resulted in large eddy fluxes close to the coast on the
western side of the basin where the mean meridional
temperature gradients are large. However, as shown by
Swanson and Pierrehumbert (1997) the effective eddy
diffusivity in the Pacific storm tracks (diagnosed from
NCEP data) increases toward the east as storms develop,
consistent with a maximum impact of the eddies offshore. Pure isobaric mixing along a mean temperature
gradient, on which the parameterization in previous
model versions was based, would result in unrealistically small mixing lengths and damping times. The reason for the failure of the local mixing parameterizaton
could be due to the fact that mixing occurs along isentropes instead of isobars or to strong thermal interaction with the underlying oceanic surface. The latter
process can maintain the strong meridional temperature
gradient against dissipation by damping the air temperature fluctuations. This effect should be stronger in
regions of deep mixed layers (large heat capacity), that
is, just south of the Kuroshio close to the Japanese coast.
Although cooling is substantial, it is not enough to
remove all errors in the surface heat fluxes in the model.
Figure 5b shows the difference between the latent and
sensible heat fluxes in the AML model and the NCEP–
NCAR reanalysis dataset. Away from the midlatitudes,
the errors are less than 20 W m22 . Considering the uncertainties in the data, this is acceptably small. However,
large errors in the Kuroshio region remain. In winter
the outgoing heat fluxes are about 25% too low, although
the annually averaged error is only 30 W m22 in this
VOLUME 31
region. As the effect of the eddies are well represented
(Fig. 4), these errors must be due to another physical
process or due to uncertainties in the data. To explore
the latter possibility, we performed a run with the AML
model using input fields from a coupled atmospheric
general circulation (Geophysical Fluid Dynamics Laboratory) mixed layer model. (i.e., the SST and the winds
from the coupled model are specified; M. Alexander
1999, personal communication). The fluxes produced
by the AML model are compared with those from the
coupled model. The advantage of this setup is that the
surface fluxes from the coupled model are consistent
with the input fields of the AML. In this case the data
is ‘‘perfect,’’ that is, all fields are consistent with each
other. The errors in the simulated heat fluxes were virtually the same. Apparently, the AML misses an essential process in the western boundary current region. This
confirms the results of the budget analysis of the surface
layer described in section 2. The residual diagnosed in
section 2 is balanced by entrainment and radiative cooling (see Seager et al. 1995). When the parameterizations
for radiation and entrainment used in the AML model
were applied to the observed fields in order to close the
budget, a residual still remained. This residual appeared
to have the same pattern and amplitude as seen in Fig.
5. This means that all terms are modeled correctly except
for the entrainment closure (radiative cooling is uniform
and smaller). As the boundary layer air advects offshore
a deep cloud layer forms above the well-mixed subcloud
layer (e.g., Grossman and Betts 1990) as a result of
powerful shallow convection. This convection, possibly
augmented by deep frontal convection, acts to bring dry
air down from above and increases the surface fluxes.
This process is not well accounted for in the AML model. However, the goal of this study is to investigate the
impact of the storm tracks on the ocean. It is clear that
the present setup is suitable for this in that storm tracks
are well represented. The effect of the storms on the
surface fluxes has been shown to be substantial. Their
impact on the oceanic state is discussed in the next
section.
b. Impact on SST
The SST is determined by oceanic heat advection,
mixing, and surface heat fluxes. As shown in the previous sections, storms act to increase the outgoing surface heat fluxes. The ocean model coupled to the AML
has been run for 20 years with and without the parameterization of the effect of storms on the surface fluxes
to estimate the impact of this increased cooling on SST.
In Fig. 6a we show the difference between the SSTs
in January in both runs. As expected, the effect is largest in the midlatitudes. The impact of the storms is to
cool the SST by at least 3 K. The cooling is larger in
the eastern side of the basin because the mixed layer
is shallower there. Figure 6b shows the difference between SST from the NCEP–NCAR reanalysis data and
FEBRUARY 2001
HAZELEGER ET AL.
625
FIG. 5. (a) Difference between latent plus sensible heat fluxes (W m 22 ) with and without parameterization
of additional cooling due to storms in January. (b) Difference between observed (NCEP–NCAR reanalysis)
and modeled (AML) sensible plus latent heat fluxes in Jan (W m 22 ).
the modeled SST in the model including the storm
tracks. The error in SST is generally small. Without
the eddy-induced transfer velocity the SST was too
high in almost the entire basin. SST was up to 4 K too
warm in the Northern Hemisphere winter in the Ku-
roshio extension. Now the error is generally smaller
than 1 K. The error in the South Pacific has also been
reduced significantly.
Between 1508E and 1808 along 408N the simulated
SSTs are too high. This is likely associated with the too
626
JOURNAL OF PHYSICAL OCEANOGRAPHY
VOLUME 31
FIG. 6. (a) Difference between SST (K) in the LOAM model with and without parameterization of additional
cooling due to storms in Jan. (b) Difference between observed (NCEP) and modeled (LOAM) SST in Jan.
(c, d) Same as (a, b) but for Jul.
small outgoing heat fluxes produced by AML in this
region (see Fig. 5b). The SSTs are somewhat too low
along 158–258N. This may be caused by relatively high
wind speeds in this region in the NCEP–NCAR reanalysis data. Another area where the error is relatively
large is the eastern tropical Pacific. Here the error is
most likely caused by errors in the upwelling due to
surfacing of the undercurrent, a general flaw of ocean
models with relatively coarse vertical resolution. Increasing the number of layers in the upper thermocline
FEBRUARY 2001
HAZELEGER ET AL.
627
FIG. 6. (Continued )
will reduce the error (e.g., Murtugudde et al. 1996). The
change of SST in the Tropics due to the storms is remarkable since the eddy-induced transfer velocity is almost zero in this region. As the change cannot be locally
forced, a change in ocean dynamics is implicated. A
likely explanation is that changes in the subtropical meridional overturning cell, which has a downward branch
at about 308N and upwells near the equator, influences
the tropical SST (see section 5c). The presence of such
cells have been shown in models and data (e.g., Rothstein et al. 1998; Johnson and McPhaden 1999). Additionally, cooler midlatitude air is advected into the
tropical regions.
The effect of the parameterization is similar in the
628
JOURNAL OF PHYSICAL OCEANOGRAPHY
austral winter as in boreal winter (Fig. 6c). Storms act
to cool almost the entire Pacific by as much as 3 K.
Figure 6d shows that in austral winter the simulation of
SST is excellent in the South Pacific when the parameterization of cooling by the storms is included. In the
North Pacific, however, the cooling appears to be too
large, although the error is reduced. The too strong cooling is not merely caused by the storms. As mentioned
before, the NCEP–NCAR reanalysis wind speed in the
model simulation has been enhanced by a factor 1.25.
This was motivated by the low NCEP–NCAR reanalysis
wind speeds compared to European Centre for Medium
Range Weather Forecasts reanalysis and COADS winds.
The correction led to better simulated SSTs in the central
North Pacific, but it has led to too much cooling in some
other regions.
It is clear that the storms cool the SST significantly
in the midlatitudes. In general the SST is much better
simulated when the cooling by the storm tracks is included. During the winter season especially, SST errors
are greatly reduced. The cooler SSTs and increased outgoing heat fluxes affect the water mass transformation,
the mixed layer depth and the characteristics of the subducted water masses. This will be investigated in the
next section.
ON WATER MASS TRANSFORMATION
Buoyancy fluxes cause water to change its density.
The rate at which water is transformed from one density
to another by buoyancy fluxes can be determined from
the surface distribution of the surface fluxes (Walin
1982; Speer and Tzipermann 1992):
a
(E 2 P)
Q net 2 bS
.
rcp
(1 2 S/1000)
F(s) 5
1
dr
E
f (s) dA(s).
(6)
The divergence of the transformation (in density space)
is the formation of a water mass. Over the Pacific Ocean
basin, the net formation of water masses must be balanced by either an inflow or an outflow of a water mass
at the boundaries. Additionally, the diffusive mixing in
the mixed layer and in the interior contributes to the
transformation of water masses. However this plays a
minor role in water mass transformation rates in the
midlatitudes (Nurser et al. 1999; Marshall et al. 1999).
An interesting and useful diagnostic is the total diapycnal transfer by surface fluxes through surface outcrops south (or north) of a latitude Q (see Marsh et al.
2000, hereafter MNMN). This represents an overturning
streamfunction in density coordinates if the convergence
of the interior diapycnal density flux is zero and if the
volume at densities lower than s is not inflating or deflating (see MNMN for details); that is,
E
u
F du.
(7)
umax
Horizontal variations in the buoyancy fluxes induce
horizontal gradients in the mixed layer depth. Since the
amount of subducted water is significantly influenced
by lateral flow through the sloping winter mixed layer
base, the buoyancy fluxes strongly influence subduction
(Woods 1985). In addition, both the SST and sea surface
salinity are modified by heat and freshwater fluxes. Thus
the amount and the characteristics of the subducted water, and therefore the stratification, are strongly influenced by the surface buoyancy fluxes. As shown in the
previous sections, storms enhance outgoing surface heat
fluxes over the midlatitude Pacific. In the following, we
investigate the impact of the storms on the interior of
the ocean by studying transformation rates of water at
the surface, the influence of storms on mixed layer
depth, as well as effects on subsurface properties as
revealed by cross-section plots.
f (s) 5
haline expansion coefficients. The total transformation
per density class (dr) can be obtained by integrating the
transformation rates over the area (A) of the surface
outcrop of water in that density class:
c (s, u ) 5
c. Impact on the thermocline
1) IMPACT
VOLUME 31
(5)
Here Qnet is the net surface heat flux, E 2 P the freshwater flux, S the salinity, and a and b the thermal and
Here c(s, u) 5 # x:u(x)5Q # rrmax hy dr dl. This is the total
transport of water denser than r across the line of latitude u 5 Q, where rmax is the maximum density at that
latitude. Here, we will use these relations to show the
impact of the storm tracks on the heat fluxes, and hence
on the meridional overturning streamfunction.
In Fig. 7 the annual mean meridional transformation
streamfunction is presented according to Eq. (7). Positive values indicate a diapycnal volume flux toward
higher densities. For reference, the potential density at
the surface is shown in Fig. 8. The model shows a
maximum flux toward higher densities at around s 5
23 kg m23 in the North Pacific. This transport is around
28 Sv, and is in excellent agreement with observations.
Almost 50% of the strength of the shallow overturning
in the midlatitude North Pacific is generated by the cooling due to the storms (see Fig. 7b). The downward
branch of this positive cell is located at s 5 25.5 kg
m23 in the model. The outcrop of this isopycnal is just
south of the Kuroshio Extension. At this latitude, the
subtropical and central mode waters are formed (Suga
and Hanawa 1995). Indeed, the core of these homogeneneous subsurface water masses is found at s 5
25.5 kg m23 . It is clear that the storms contribute greatly
to the formation of these mode waters, which dominate
the stratification of the upper North Pacific.
In the model, there is too much transport toward lower
densities in the tropical regions (around s 5 21.5 kg
m23 ). Storms act to reduce the strength of this tropical
overturning cell by as much as 30 Sv, but this is still
FEBRUARY 2001
HAZELEGER ET AL.
629
FIG. 7. Annually averaged transformation streamfunction (Sv). (a) Model (LOAM) with parameterization of additional cooling due to
storms, (b) difference between model (LOAM) with and without the parameterization, and (c) according to NCEP buoyancy fluxes and
Levitus SST and SSS in the North Pacific. (d), (e), (f ) As in (a), (b), and (c) but for the South Pacific.
not enough. The main reason for the too low values in
the tropical region is that the area covered by water of
s ø 21 kg m23 is lower in the model compared to the
observations. As shown in Fig. 5a the heat fluxes in the
Tropics are close to observations. Nevertheless, the im-
pact of the storms on the Tropics is remarkable. As u*
is small in this region, the impact is indirect. Apparently
a change in the shallow meridional overturning cells in
the ocean is induced by the u* (see Fig. 7b). Changes
in the downwelling branches of the overturning are ex-
630
JOURNAL OF PHYSICAL OCEANOGRAPHY
VOLUME 31
FIG. 8. Maximum potential density at the surface (kg m23 ).
pected to induce changes in the upwelling branches as
well, as interior mixing is small away from the Tropics.
However, we should be careful, as vertical mixing is
expected to contribute significantly to the overturning
in the Tropics (Nurser et al. 1999). In the present analysis the role of vertical mixing could not be included.
That the overturning is too strong at low latitudes may
be due to errors in the heat fluxes in the eastern Pacific.
Here the undercurrent surfaces, resulting in SSTs that
are too low. In response, there is too much warming,
as revealed by the heat fluxes. As stated before, an increase in the number of layers in the ocean model may
resolve some of the problems.
The lightening of surface waters around s 5 26 kg
m23 is absent in the observations and seems to be enhanced by the storms in the model. The storms act to
heat in the summer in a small region around 458N, 1608E
and in the Sea of Okhotsk (see Fig. 2b).
In the South Pacific two distinct maxima are found
in the model simulation: one at s 5 23 kg m23 and
another at s 5 24.8 kg m23 . The observations also show
maxima, although at slightly higher density classes and
less clear than in the model. The diapycnal flux toward
lower densities at high (s 5 27.2 kg m23 ) and low
densities (s 5 21.5 kg m23 ) are in agreement with observations. Again the cooling by the storms contributes
enormously to the transformation of water masses in the
midlatitudes. Without the parameterization the positive
overturning cell at s 5 24.8 kg m23 is almost absent.
Also, there is a strong lightening of waters around s 5
23.5 kg m23 without storms. Finally, the negative cell
around s 5 27.2 kg m23 is almost entirely caused by
the storms. The small negative cell around s 5 24 kg
m23 is not found in the observations although there is
a local minimum at s 5 24.5 kg m23 . The storms contribute more to the transformation of water masses in
the South Pacific than in the North Pacific. This is mainly due to the small seasonal cycle and the larger area
affected in the South Pacific.
From the transformation streamfunctions it is clear
that the storms have a large impact on the transformation
of water masses at midlatitudes. Especially the overturning cells in the density range from s 5 22 kg m23
to 25 kg m23 are affected. The downwelling branches
of these circulations correspond to the formation regions
of mode waters that cover the upper part of the thermocline in the ocean. In the next section we will study
how the interior is affected by the storms.
2) IMPACT
ON VERTICAL STRUCTURE OF THE
OCEAN
Subduction at the base of the mixed layer leads to
formation of water masses. As diapycnal mixing below
the mixed layer is small, the amount and the characteristics of the subducted water determine the vertical
stratification of the ocean. The contribution of lateral
flow through the sloping mixed layer depth in late winter
to subduction can be large (Woods 1985). In the North
Pacific, gradients in mixed layer depth are large in the
Kuroshio region. In this area the contribution of lateral
induction to the total subduction is large (Qiu and Huang
1995). The horizontal variations in the surface fluxes
induce horizontal variations in the mixed layer depth.
FEBRUARY 2001
HAZELEGER ET AL.
631
FIG. 9. Mixed layer depth (m) in March (a) model with parameterization of additional cooling due to
storms, (b) difference in model with and without the parameterization, and (c) as obtained from Levitus
data.
So, the cooling by the storms, which occurs in the region
of high mixed layer depth gradients, is expected to have
an impact on mixed layer depth and on subduction.
Indeed, the maximum mixed layer depth reaches to
250 m in the Kuroshio Extension when the eddy-induced transfer velocity is included, while without the
extra cooling by storms the mixed layer depth reaches
only to 175 m (Fig. 9). The maximum deepening due
to the cooling by eddies is 110 m. The observed max-
imum mixed layer depths reach to 250 m. The changes
in mixed layer depth take place mainly in wintertime
(Fig. 10). So, the additional eddy-induced transfer velocity leads to better simulated mixed layer depths. As
the cooling enhances the transformation overturning,
steepens mixed layer gradients, and increases late winter
mixed layer depths, more voluminous and cooler water
masses are formed when the cooling by storms is included in the model. This is clearly visible in the cross
632
JOURNAL OF PHYSICAL OCEANOGRAPHY
VOLUME 31
nous and denser water masses are formed in the midlatitudes due to the presence of the storm tracks.
Figure 12 shows the change in the overturning mass
streamfunction in a meridional cross section. The shallow overturning increases by about 1.6 Sv, a 10% increase. Associated changes in meridional velocities are
of the order of less than a centimeter per second. Zonal
velocities change by a few centimeters per second in
the equatorial region. The overturning also deepens.
These changes in mass transport are not very large.
Large changes take place in the subsurface temperature
(Fig. 13), especially in ventilated regions where cooling
of a few degrees is found. These tongues with anomalous cool water clearly originate from outcrops in the
midlatitudes. Even in the Tropics cooling of 1 K is
found. Salinity hardly changes in the Tropics but increases in the directly ventilated zones (Fig. 13a). Thus,
the impact of the cooling by the storms appears to be
mainly of thermal nature. It is these rearrangements of
density that produce the large changes in the transformation streamfunction, which produces homogeneous
subsurface water masses. The changes in the salinity
and temperature cause a realistic representation of the
characteristics of the water masses in the ventilated thermocline.
6. Conclusions
FIG. 10. Seasonal cycle of mixed layer depth (m) at 368N, 1758W
in the model with (continuous line) and without (dashed line) cooling
by the storms.
sections of potential density shown in Fig. 11. Around
308N a homogeneous water mass is found in the observations. This is the North Pacific subtropical mode
water (Suga and Hanawa 1995). The core of the homogeneous layer is observed at s 5 25.5 kg m23 . In
the model, a very distinct mode water layer is simulated
between s 5 25.6 kg m23 and 25.2 kg m23 in accordance
with the observations. Without the parameterization of
the additional cooling by storms this homogeneous water mass is not well simulated. Also, the potential density
is too low in the upper thermocline. It has been shown
already in Figs. 7a and 7b that the storms are responsible
for the positive overturning cell associated with formation of a water mass around s 5 25.5 kg m23 .
In the cross sections this conclusion is confirmed. The
mode water layer in the climatology is less homogeneous than in the model. However, one must take into
account that a climatological forcing has been used in
the model. Thus the same water mass is formed every
year. In the real world, the forcing varies. This leads to
varying characteristics of mode water, in accordance
with variations in the forcing (Hazeleger and Drijfhout
1998). Generally, one can conclude that the storms enhance the subtropical overturning cell. More volumi-
A good understanding of the surface forcing of the
ocean is of prime importance for climate and oceanographic studies. Here we studied one aspect of the forcing of the ocean, namely, the impact of the midlatitude
storm tracks. An analysis of the moist static energy
budget of the atmospheric surface layer in the Pacific
region showed that the atmospheric eddies impose a
cooling on the surface layer. This cooling occurs in a
zonal band over the extratropical North and South Pacific. The cooling is associated with the northward heat
transport in the atmosphere that is accomplished by transient eddies. The maximum cooling is found to be 60
W m22 at 1708E. The maximum cooling by the atmospheric eddies occurs far offshore. Close to the coast
the storms still develop and the surface fluxes damp the
air temperature variance, maintaining strong meridional
temperature gradients (Swanson and Pierrehumbert
1997). It is worth noting that this implies that parameterizing the eddy-induced transfer velocity (u* ) in
terms of surface air temperature gradients does not work
very well.
To study how the cooling by the storms affect the
surface and subsurface layers of the ocean, an ocean
model coupled to a model of the atmospheric mixed
layer has been used. A parameterization of the impact
of the eddies on the heat budget of the surface layer has
been implemented that is based on a parameterization
of eddies in the ocean. The storms are represented by
an eddy-induced transfer velocity that is added to the
Eulerian mean velocity. In this study the eddy-induced
FEBRUARY 2001
HAZELEGER ET AL.
633
FIG. 11. Cross section of potential density (kg m23 ) in Sep in (a) the model with and (b) without
parameterization of additional cooling due to storms and (c) according to Levitus data at 1508E.
The isopycnals where mode water is observed are dashed.
transfer velocity has been specified according to the
NCEP–NCAR reanalysis data. The simulated eddy
transports of air temperature and humidity by the atmospheric mixed layer model are in good agreement
with the observed eddy transports. To study the effect
of storm tracks on oceanic variables the coupled model
has been run with and without the eddy-induced transfer
velocity.
The enhanced heat fluxes from the ocean to the atmosphere lead to SSTs that are 3 K lower in the midlatitudes. Overall the surface characteristics of the ocean
are more realistically simulated when the cooling by the
storms is included. However, in the boreal summer a
relatively large error has been induced in the subtropical
regions. This deficiency may be attributed to errors in
the wind climatology in this region. The cooling attributable to the storms makes surface waters less buoyant,
which leads to deeper mixed layers in the midlatitudes.
This results in an increased subduction of denser water
in late winter. As a result the shallow overturning is
more intense and more voluminous water masses are
formed. The homogeneous mode waters that dominate
the stratification in the subtropical gyre are especially
well simulated when the cooling effect of storms is accounted for. As the subtropical meridional overturning
cell is enhanced by the eddies and cooler water is subducted, even the tropical regions are affected. Although
the storms do not directly affect the surface layers here,
the upwelling branch of the overturning cell connects
the midlatitudes to the (sub)tropics. As a result of the
634
JOURNAL OF PHYSICAL OCEANOGRAPHY
VOLUME 31
FIG. 12. Zonally averaged overturning streamfunction (Sv) in (a) model with cooling due to the
storms and (b) the difference in the model with and without the additional cooling.
change in the meridional cell, even tropical SSTs are
affected. This mechanism leaves open the possibility of
midlatitude variability influencing tropical variability
via an oceanic bridge, as has been suggested by Gu and
Philander (1997).
In conclusion, the cooling by the atmospheric storm
tracks contributes significantly to the heat budget of the
midlatitude atmospheric surface layer. Its effect on the
surface and subsurface layers of the ocean has been
shown to be substantial. As such, a parameterization of
the impact of atmospheric storms on the surface heat
budget is essential when ocean models are coupled to
simple atmospheric models that do not include the impact of transient eddies. The effect of the atmospheric
FEBRUARY 2001
HAZELEGER ET AL.
635
FIG. 13. Meridional cross section of impact of the cooling by storms on (a) the salinity (psu) and (b)
temperature (K).
eddies can be represented well by an eddy-induced
transfer velocity. The next challenge will be to develop
a parameterization of turbulent entrainment at the top
of the atmospheric mixed layer. In addition, there is also
the challenge of relating variability in the storm tracks
to that of the underlying SST. We plan to study these
issues in the near future.
Acknowledgments. The authors would like to thank
Benno Blumenthal for pioneering work on the subject
of this paper. This work was sponsored by NOAA
(UCSIO-10775411D/NA47GP0188, ‘‘The Consortium
on the Ocean’s Role in Climate’’) and NSF (ATM-9224915). W.H. likes to thank the people of the Climate
and the Physical Oceanography group of Lamont Doherty Earth Observatory for their hospitality and stimulating atmosphere during his stay. Furthermore, KNMI
is thanked for the opportunity and financial support of
the visit.
REFERENCES
Alexander, M. A., and S. D. Schubert, 1990: Regional earth–atmosphere energy balance estimates based on assimilations with a
GCM. J. Climate, 3, 15–31.
, and J. D. Scott, 1997: Surface flux variability over the North
Pacific and North Atlantic Oceans. J. Climate, 10, 2963–2978.
Bishop, J. K. B., and W. B. Roswow, 1991: Spatial and temporal
variability of global surface solar irradiance. J. Geophys. Res.,
96, 16 839–16 858.
Boers, R., and A. K. Betts, 1988: Saturation point structure of marine
stratocumulus clouds. J. Atmos. Sci., 45, 1156–1175.
Camp, N. T., and R. L. Elsberry, 1978: Oceanic thermal response to
strong atmospheric forcing. II: The role of one-dimensional processes. J. Phys. Oceanogr., 8, 215–224.
Cayan, D., 1992: Latent and sensible heat flux anomalies over the
northern oceans: Driving the sea surface temperature. J. Phys.
Oceanogr., 22, 541–560.
Da Silva, A. M., C. C. Young, and S. Levitus, 1994: Atlas of Surface
Marine Data 1994, Vol. 1: Algorithms and Procedures, NOAA/
NESDIS Atlas 6, U.S. Department of Commerce, NOAA, NESDIS, 83 pp.
Gent, P. R., and J. C. McWilliams, 1990: Isopycnal mixing in ocean
circulation models. J. Phys. Oceanogr., 20, 150–155.
Grossman, R. L., and A. K. Betts, 1990: Air–sea interaction during
an extreme cold air outbreak from the eastern coast of the United
States. Mon. Wea. Rev., 118, 324–342.
Gu, D. F., and S. G. H. Philander, 1997: Interdecadal climate fluctuations that depend on exchanges between the Tropics and extratropics. Science, 275, 805–807.
Hazeleger, W., and S. S. Drijfhout, 1998: Mode water variability in
a model of the subtropical gyre: Response to anomalous forcing.
J. Phys. Oceanogr., 28, 266–288.
Johnson, G., and M. McPhaden, 1999: Interior pycnocline flow from
the subtropical to the equatorial Pacific Ocean. J. Phys. Oceanogr., 29, 3073–3089.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.
Kleeman, R., and S. B. Power, 1995: A simple atmospheric model
of surface heat flux for use in ocean modeling studies. J. Phys.
Oceanogr., 25, 92–105.
Kushner, P. J., and I. M. Held, 1998: A test, using atmospheric data,
of a method for estimating oceanic eddy diffusivity. Geophys.
Res. Lett., 25, 4213–4216.
Latif, M., and T. P. Barnett, 1994: Causes of decadal climate vari-
636
JOURNAL OF PHYSICAL OCEANOGRAPHY
ability over the North Pacific and North America. Science, 266,
634–637.
Lau, N., 1988: Variability of the observed midlatitude storm tracks
in relation to low-frequency changes in the circulation pattern.
J. Atmos. Sci., 45, 2718–2743.
Levitus, S., and T. P. Boyer, 1994. World Ocean Atlas 1994, Vol. 4:
Temperature. NOAA Atlas NESDIS, U.S. Govt. Printing Office,
117 pp.
, R. Burgett, and T. P. Boyer, 1994: World Ocean Atlas 1994,
Vol. 3: Salinity. NOAA Atlas NESDIS, U.S. Govt. Printing Office,
99 pp.
Marsh, R., A. J. Nurser, A. P. Mergann, and A. L. New, 2000: Water
mass transformation in the Southern Ocean of a global isopycnal
coordinate GCM. J. Phys. Oceanogr., 30, 1013–1045.
Marshall, J. C., D. Jamous, and J. Nilsson, 1999: Reconciling ‘thermodynamic’ and ‘dynamic’ methods of computation of water
mass transformation rates. Deep-Sea Res., 46, 545–572.
Murtugudde, R., R. Seager, and A. J. Busalacchi, 1996: Simulation
of the tropical oceans with an ocean GCM coupled to an atmospheric mixed layer model. J. Climate, 9, 1795–1815.
Naik, N. H, M. A. Cane, S. Basin, and M. Israeli, 1995: A solver for
the barotropic mode in the presence of variable topography and
islands. Mon. Wea. Rev., 123, 817–832.
Nurser, A. J. G., R. Marsh, and R. G. Williams, 1999: Diagnosing
water mass transformation from air–sea fluxes and surface mixing. J. Phys. Oceanogr., 29, 1468–1487.
Polonsky, A. B., S. A. Baev, and A. Diakite, 1992: On the response
of the ocean upper layer to synoptic variability of the atmosphere. Dyn. Atmos. Oceans, 16, 225–248.
Qiu, B., and R. X. Huang, 1995: Ventilation of the North Atlantic
and North Pacific: Subduction versus obduction. J. Phys. Oceanogr., 25, 2374–2390.
Rodgers, K. B., M. A. Cane, N. H. Naik, and D. P. Schrag, 1999:
The role of the Indonesian Throughflow in equatorial Pacific
thermocline ventilation. J. Geophys. Res., 104, 20 551–20 570.
VOLUME 31
Rothstein, L. M., R. Zhang, A. J. Bussalacchi, and D. Chen, 1998:
A numerical simulation of the mean water pathways in the subtropical and tropical Pacific Ocean. J. Phys. Oceanogr., 28, 322–
343.
Seager, R., M. B. Blumenthal, and Y. Kushnir, 1995: An advective
atmospheric mixed layer model for ocean modeling purposes:
Global simulation of surface heat fluxes. J. Climate, 8, 1951–
1964.
, Y. Kushnir, M. Visbeck, N. Naik, J. Miller, G. Krahmann, and
H. Cullen, 2000: Causes of Atlantic Ocean climate variability
between 1958 and 1998. J. Climate, 13, 2845–2862.
Speer, K., and E. Tziperman, 1992: Rates of water mass formation
in the North Atlantic Ocean. J. Phys. Oceanogr., 22, 93–104.
Suga, T., and K. Hanawa, 1995: The subtropical mode water circulation in the North Pacific. J. Phys. Oceanogr., 25, 958–970.
Swanson, K. L., and R. T. Pierrehumbert, 1997: Lower-tropospheric
heat transport in the Pacific storm track. J. Atmos. Sci., 54, 1533–
1543.
Valdes, P., and B. J. Hoskins, 1989: Linear stationary wave simulations of the time-mean climatological flow. J. Atmos. Sci., 46,
2509–2527.
Visbeck, M., J. Marshall, T. Haine, and M. Spall, 1997: Specification
of eddy transfer coefficients in coarse resolution ocean circulation models. J. Phys. Oceanogr., 27, 381–402.
, H. Cullen, G. Krahmann, and N. Naik, 1998: An ocean models
response to North Atlantic Oscillation-like wind forcing. Geophys. Res. Lett., 25, 4521–4524.
Walin, G., 1982: On the relation between sea-surface heat flow and
thermal circulation in the ocean. Tellus, 34, 187–195.
Wallace, J. M., G. H. Lim, and M. L. Blackmon, 1988: Relationship
between cyclone tracks, anticyclone tracks, and baroclinic wave
guides. J. Atmos. Sci., 45, 439–462.
Woods, J. D., 1985: The physics of thermocline ventilation. Coupled
Ocean–Atmosphere Models, J. C. J. Nihoul, Ed., Elsevier, 543–
590.